Uploaded by Jacob G

508 UEPC Chronic Absenteeism Research Brief

advertisement
Contents
July 2012
Characteristics of Chronically Absent
Students
2
Research Brief:
Chronic Absenteeism
Who are the Chronically Absent
Students ?
6
CA Across School Years
8
The CA-Dropout Relationship
10
This research brief focuses on Chronic Absenteeism (CA) in Utah public schools. We address:

The students who are most likely to be chronically absent

A demographic profile of chronically absent students

Patterns in chronic absenteeism over time

Relationships between chronic absenteeism and lower standardized test scores

Relationships between chronic absenteeism and dropping out

The extent to which grade point average
(GPA) mediates the relationship between
chronic absenteeism and dropping out
student is chronically absent if he or she
misses school 10 percent of the time, or
more, for any reason, according to Attendance
Works. (Attendance Works is a national initiative
that promotes awareness of attendance issues. See
http://www.attendanceworks.org/) Researchers
have identified chronic absenteeism as a persistent
problem related to poor academic performance and
potential behavioral and developmental issues.
There is general agreement among researchers that
being chronically absent places students at risk of
negative academic consequences (Chang &
Romero, 2008; Moonie, Sterling, Figgs, & Castro,
2008).
A small but growing body of research based on
chronic absenteeism data has emerged. Recent
research indicates that:

Chronic absenteeism in kindergarten can
be negatively correlated with academic
What is chronic absenteeism?
A student is chronically absent if
he or she misses school 10
percent of the time, or more, for
any reason.
performance in the first grade (Chang &
Romero, 2008).

Chronic absenteeism can have pronounced
negative impacts on students of poverty
(Ready, 2010).

Chronic absenteeism is often higher in
urban, as compared to rural, schools
(Sheldon & Epstein, 2004).

Chronic absenteeism can be an early
predictor of dropping out of high school
(Mac Iver & Mac Iver, 2010).

Chronic absenteeism can reduce the
likelihood of post-secondary enrollment
(Balfanz & Byrnes, 2012).
This report brief and the research conducted herein were completed by the Utah Education Policy Center. For more
information about this study or the Utah Education Policy Center, visit the Center’s Web site at http://uepc.ed.utah.edu.
Although No Child Left Behind compels states to
report attendance, there are no mandates to report
chronic absenteeism. As a result, most states and
local education agencies report attendance rates to
meet reporting requirements (Balfanz & Byrnes,
2012). Unfortunately, reporting average attendance
rates can obscure the number of chronically absent
students. For example, a school with 500 students
and a 94 percent attendance rate could have from
zero to 250 chronically absent students, depending
on how the absences are distributed. A primary
consequence of only reporting or attending to
attendance rates is that supports and services for
students who are chronically absent may be
limited.
Utah provides a more specific example. Overall,
students enrolled in Utah public schools attended
approximately 95 percent of the days for which
they were enrolled. A 95 percent average
attendance rate seems encouraging. However, the
95 percent average attendance rate obscures the
fact that 13.5 percent of all Utah students were
chronically absent during that same year. As is
generally the case, reporting only attendance
averages does not consider the number of students
who were chronically absent.
Encouragingly, six states—Georgia, Florida,
Maryland, Nebraska, Oregon and Rhode Island—
now collect and report chronic absenteeism rates.
For the 2010-2011 academic year, those six states
reported chronic absenteeism rates ranging from a
low of 6% to a high of 23% of students chronically
absent statewide (Balfranz & Byrnes, 2012).
This study expands the existing research cited
above on chronic absenteeism by including Utah in
the growing number of states that have analyzed
their chronic absenteeism data. This research brief
also extends the body of research on chronic
absenteeism by statistically examining Grade Point
Average (GPA) as a mediating variable in the
relationship between chronic absenteeism and
dropping out.
Characteristics of Chronically Absent
Students
In order to study the effects of chronic absenteeism
in Utah, we constructed two data sets for this
report. One was cross-sectional and included all
Utah public school students enrolled in the 20102011 school year.1 The other was longitudinal and
followed the class of 2010 for five years, from
their 8th grade year in 2006 through their
graduation year in 2010.2
We used the cross-sectional data set to examine
relationships between chronic absenteeism and
three categories of student variables: predictor
variables, covariates, and outcome variables.
Predictor variables, covariates, and outcome
variables were identified from the literature and
were defined, and used, as follows:
1. Characteristics that predict chronic
absenteeism: We used Low Income,
Special Education, English Proficiency
and Racial Minority as variables to predict
chronic absenteeism because their values
were known at the beginning of the year,
prior to the occurrence of the absences.
These variables were recorded at the
beginning of the academic year and were
not changed over the course of the year.
2. Characteristics that co-occur with
chronic absenteeism: Other variables
could have been recoded, or updated,
during the year. We used Mobility and
Homelessness as covariates. These two
variables could not be considered as either
predictor nor outcome variables because
there was no way of knowing whether any
given student was first chronically absent
and then mobile or first mobile and then
chronically absent (or if the two events cooccurred).
3. Characteristics that are outcomes of
chronic absenteeism: Finally, some
Page | 2
variables were measured at the end of the
year. We used Reading on Grade Level,
CRT Scores, Cumulative GPA and
Dropping Out as outcome variables
because they were measured after the
absences occurred and could have been
affected by student attendance (but not
vice-versa).
because both Utah boys and girls were chronically
absent at the exact same rate: 13.5 percent. Other
studies have shown the same lack of effect for
gender (e.g., Atwood and Croll, 2006). Figure 2
shows the change in odds associated with
membership in any of the seven race categories
reported by the Utah State Office of Education
(USOE).
Predictors of Chronic Absenteeism
Chronic absenteeism was predicted by the four
variables identified as predictor variables (i.e.,
Low Income, Special Education, English
Proficiency and Racial Minority). Results are
reported as change in odds ratios: a commonly
reported effect size for research with “yes or no”
outcomes (in this case chronically absent or not). 3
In general, a change in odds of one indicates the
exact same outcome for members and nonmembers of the group being analyzed. Odds ratios
greater than one indicate that members of the
group being analyzed have odds of the outcome (in
this case odds of being chronically absenteeism)
that are increased that many times compared to
non-members of that group. For example, the
change in odds statistic related to low income is
1.9. This indicates that a student who received free
or reduced lunch was 90 percent (1.9 times) more
likely to be chronically absent than a student who
does not receive free or reduced lunch.
Figure 1shows the change in odds, or likelihood, of
being chronically absent given membership in each
of the predictor categories (i.e., Race, English
Proficiency, SES, and Special Education)
compared with non-membership in those
categories. Notably, sex is not reported. This is
A student who received free or
reduced lunch was 90 percent more
likely to be chronically absent than
a student who did not receive free or
reduced lunch.
Figure 1. Increased Odds of Chronic Absence Given
Membership in a Predictor Group
Change in Odds of Chronic Absenteeism
Racial Minority
LEP
Special Education
Low Income
1.4
1.2
1.7
1.9
These results show the odds of being chronically
absent associated with each of the predictor
variables (i.e., income, special education, English
Language Learner, and Race). Students from all
of the groups represented in Figure 1 (i.e., racial
minority, LEP, special education and low income)
had significantly higher odds of being chronically
absent than their peer students not categorized into
those groups. The increased odds of chronic
absence were highest for students from low income
homes (about 90 percent higher than for students
not from low income homes ). There was a small
difference (about a 20 percent increase in odds)
between students identified as not yet English
proficient and their English proficient peers.
Page | 3
Covariates and Chronic Absenteeism
The covariates, Mobility (measured as a student
having been unenrolled from at least one Utah
school and re-enrolled into another during the
course of a school year) and Homelessness were
considered in the same way as the predictor
variables were (i.e., change in odds). As seen in
Figure 3, the change in odds of being chronically
absent were quite a bit higher for the covariates
(i.e., mobility and homelessness) than for the
demographic predictors used in the previous
section.
Figure 2. Change in Odds of Chronic Absence Given
Membership in Any of the USOE Racial Categories
Multi-racial
1.17
White
0.7
Pacific Islander
1.38
Hispanic
1.4
Black
1.34
Asian
0.61
American Indian
0
0.5
2.13
1
1.5
2
2.5
The relationship between homelessness and
absenteeism is an issue that has been addressed
nationally by the McKinney-Vento act. The
McKinney-Vento Act (established in 1987 and
reauthorized in 2001) requires states to implement
measures to eliminate school enrollment barriers
(e.g., residency requirements, documentation and
immunization records, guardianship issues, and
lack of uniforms or appropriate clothing for dress
codes) faced by homeless students (U.S.
Department of Education, 2002, U.S. Department
of Education, 2004). In 2001, research in New
York Schools that used attendance rates, not
chronic absenteeism rates as the outcome variable
demonstrated that homeless students were no
more likely to be absent from school than students
who were from low-income homes but were not
homeless (Buckner, Bassuk & Weinreb, 2001). In
comparison, we found in our analysis that students
who were categorized as homeless were 80 percent
more likely to be chronically absent than their low
income peers who were not homeless. Figure 4
shows the proportion of chronically absent students
from the six homeless categories reported by the
USOE. Because of the small numbers of students
in some of these categories, the results for the
homelessness categories are presented as
percentages, not the change in odds statistic
reported in the other tables.
Change in Odds of Chronic Absenteeism
Note. These results are presented on a grid that centers
on one. One indicates no change in odds, or a 1:1 ratio.
Values less than one indicate that membership in those
racial categories reduced, rather than increased, the
likelihood of a student being chronically absent.
Mobile students who moved in and out of schools
were four times more likely than non-mobile
students to be chronically absent. Students who
were homeless during the 2010-2011 school year
were two and a half times more likely to be
chronically absent than students who were not
homeless.
Figure 3. Change in Odds Associated with the
Covariates of Mobility and Homelessness
4.2
2.5
Homeless
Mobile
Change in Odds of Chronic Absenteeism
Page | 4
Figure 4. Proportion of Students from Each Homelessness Category who were Chronically Absent
38.42
39.6
29.17
29.54
27.04
23.74
12.12
Not homeless
With another
family due to
loss of housing
or economic
hardship
In a motel or
hotel
In a shelter
In a car, park,
campground or
public place
Somewhere Unaccompanied
without
minor
adequate
facilities
Proportion of Students who were Chronically Absent
Note. These categories of homelessness are used and reported by the USOE in their annual Point-in-Time count
(USOE, 2011). The USOE and the U.S. Department of Education differ slightly in their definitions of homelessness,
with the USOE counting as homeless the families or children living with another family due to hardship.
Outcome Variables
All four variables identified as outcome variables
were predicted by chronic absenteeism. The
outcomes variables were Reading Proficiency
(grades 1-3), CRT scores (grades 3-12),
Cumulative GPA scores (grades 9-12), and
Dropout (Any grade). All outcomes correlated
significantly with chronic absenteeism and in all
cases negative outcomes were associated with
chronic absenteeism. Table 1 shows the 2011
academic outcome variables and the effect that
chronic absenteeism had on those variables.
Students who were homeless
during the 2010-2011 school year
were two and a half times more
likely to be chronically absent than
students who were not homeless.
Table 1. Relationship between Chronic
Absenteeism and Academic Outcome Variables
Outcome
Effect of Chronic Absence
Reading on
grade level
CRT
Language
CRT Math
Odds of being below grade level were
1.7 times higher
Decreased 3.798 points, on average
CRT Science
Decreased 4.850 points, on average
Cumulative
GPA
Dropout
Decreased .854 points, on average
Decreased 5.861 points, on average
Odds of dropping out were 7.4 times
higher
Note. Dropout statistics in this table are based the
longitudinal data set and estimated using survival
analysis. All other statistics are based on the crosssectional data set and estimated through simple logistic
regression.
Page | 5
These outcomes showed the pervasive negative
academic influence of chronic absence across all
grade levels and in all tested subjects. The strong
relationship between chronic absenteeism and
dropping out of school is discussed in detail in a
latter section of this brief.
Who are the Chronically
Absent Students?
When working to prevent chronic absences, it is
important to understand who is at risk of
becoming chronically absent. Moreover, when
developing strategies to help students who are
already chronically absent, it is important to
understand the characteristics of those students,
regardless of risk. The purpose of this section is
to describe the demographic characteristics of
students who were chronically absent in 2011.
Over-representation of the more chronically
absent groups changes the profile from that of the
general population. Yet, with the exception of
students from low-income homes, the groups
representing the vast number of students (i.e.,
students who were white, English proficient, nonmobile, and not receiving special education
A student who was chronically
absent in any year, starting in the
8th grade, was 7.4 times more
likely to drop out of school than a
student who was not chronically
absent during any of those years.
services) are the majority of the chronically absent
students. This may seem counter-intuitive that the
groups with the lowest risk for chronic
absenteeism, as demonstrated earlier, make up the
majority of chronically absent students. However,
because of differences in the number of students
who are in the groups, this is possible. For
example, while only 5.3% of the students who
were chronically absent were homeless, students
in homeless situations were more than twice as
likely to be chronically absent than their peers
who were not homeless. This is because so few
Utah students were homeless (2.6 percent in
2011). Figure 5 shows the percentage of
chronically absent students from each of six
demographic categories.
Although over-representation of the more chronically
absent subgroups changes the profile from that of the
general population, it is important to note that, with the
exception of students from low-income homes, the
majority groups (i.e., students who were white, English
proficient, non-mobile, and not receiving special
education services) represented the majority of the
chronically absent students.
Page | 6
Figure 5. Percentage of Chronically Absent Students from Different Demographic Categories
Non-mobile 66%
Mobile
34%
Race
Other 8%
Hispanic
20%
White 72%
Mobility
Abenteeism Profile
Page | 7
Chronic Absenteeism Across
School Years
Grade Level in School
When we looked at the percent of chronically
absent students by school year, we found that
kindergarten and first grade students tended to be
chronically absent more often than their older
elementary school peers (i.e., second through
sixth graders). Once in junior high school, the
chronic absenteeism rates began to rise, increasing
each year to a peak of 20.1 percent of students
chronically absent during their senior year (See
Figure 6). This is the exact pattern of absence
reported in a recent study conducted in Oregon
(ECONorthwest, 2011) and fits the general pattern
seen across the country (Balfanz and Byrnes,
2012) and internationally (Attwood and Croll,
2006).
Figure 6. Proportion of Chronically Absent
Students by Year in School
Twelth
20.1
Eleventh
19.2
Tenth
16.5
Ninth
14.7
Eigth
Repeated Chronic Absenteeism
Other research has shown that high truancy rates
(truancy rates include only unexcused absences,
whereas chronic absence rates include both
excused and unexcused absences ) in one year of
school significantly predicted high truancy rates in
another year of school (Attwood and Croll, 2006,
Sheldon and Epstein, 2004). This pattern was
seen in the analysis of Utah data, using chronic
absenteeism instead of truancy, as well. Of the
35,508 students in our longitudinal data set, 9,847
(27.7 percent) were chronically absent at least one
year between the 8th and 12th grades. Of those
9,847 students who were chronically absent at
least once, 5,015 (51 percent of the chronically
absent students and 14 percent of all students)
were chronically absent in more than one year.
This shows that chronic absenteeism is not an
isolated event usually. To explore this
relationship more thoroughly, we ran a series of
logistic regressions that used chronic absenteeism
in one year to predict chronic absenteeism in the
subsequent year. Results showed that the
likelihood of being chronically absent in any
school year increased anywhere from 8 to 17
times (depending on the year) if the student had
been chronically absent in the previous school
year. These results are presented in Table 2.
14.2
Seventh
11.6
Sixth
10.6
Fifth
10.2
Fourth
9.8
Third
9.8
Second
On average, being
chronically absent in one
grade increased the odds
of being chronically
absent in the next grade
by nearly thirteen times
10.8
First
12.1
K
16
0
5
10
15
20
25
Page | 8
Table 2. The Increase in Odds of Being Chronically Absent in One Grade, Given Chronic Absenteeism in the
Previous Grade
Being Chronically
Absent in Grade
8
9
10
11
Odds of Being
Chronically Absent
in Grade 9
Odds of Being
Chronically Absent
in Grade 10
Odds of Being
Chronically Absent
in Grade 11
Odds of Being
Chronically Absent
in Grade 12
17.3 times
13.3 times
Our findings show that if a student was
chronically absent in one school year, his or her
chances of being chronically absent in the next
year increased thirteen-fold, on average. There
was an interesting trend in these results, wherein
the ability to predict subsequent chronic
absenteeism was greater in the earlier grades (i.e.,
eighth and ninth) than it was in the later grades
(i.e., tenth through twelfth). Our ability to predict
subsequent absenteeism based on absenteeism in
earlier grades is an important finding because, as
the next section of this report will show, the
negative impacts of chronic absenteeism are
cumulative. In future research, we hope to
examine this relationship in elementary school
and middle school students.
Chronic Absenteeism and Dropout
Over Time
Chronic absenteeism and dropout may co-occur.
Students who dropout during the school year may
be chronically absent as a part of the dropping-out
process. That is, they gradually attend class less
and less until they “officially” drop out. In these
cases, when dropout and chronic absence cooccur, it is not appropriate to think of chronic
absence as a predictor and dropout as an outcome.
In a previous analysis presented in the outcome
variables section, we did not make a distinction
between when the chronic absenteeism occurred
in relationship to when students dropped out.
However, given our understanding that chronic
12.6 times
8.1 times
absenteeism and dropping out may co-occur, it
becomes important to consider how chronic
absence in one year might predict dropping out in
a future year. Our previous analysis showed that
a student who was chronically absent in any year,
starting in the 8thgrade, was 7.4 times more likely
to drop out of school than a student who was not
chronically absent during any of those years.
However, that analysis did not necessarily help us
predict dropping out in future years based on
chronic absence in previous years. Therefore, to
control for the expected chronic absenteeism in
the dropout year, we ran the analysis again but
excluded attendance data from the last year that
the student attended school. Using the new
predictor (i.e., chronic absence in the years prior
to dropout), we were able to look at the odds of a
chronically absent student dropping out of school
in a future school year. The results showed that
students who were chronically absent were 5.5
times more likely to dropout in a future year than
their non-chronically absent peers.
We analyzed the relationship between chronic
absenteeism and dropping out in a future year by
specifying hazard functions for students who were
and were not chronically absent.4 The hazard
functions are illustrated in Figure 7, in which the
probability of a student dropping out in any grade
is plotted for both students who were, and
students who were not, chronically absent in a
year prior to the final year.
Page | 9
Figure
7. Probabilities
of Dropout
at at
Each
andWere
WereNot
NotChronically
Chronically
Figure
7. Probabilities
of Dropout
EachSchool
SchoolYear
Yearfor
forStudents
Students who
who Were
Were and
Absent
Absent
Probability of Dropout
0.3
0.25
0.2
Not Chronically Absent prior to
final year
0.15
Chronically Absent prior to final
year
0.1
0.05
0
8th
9th
10th
11th
Figure 7 illustrates several important findings.
First, there was an exponential increase in the risk
of dropping out as students approach their
scheduled graduation. The data also showed
(Figure 7) the difference in dropout outcomes for
students who were and were not chronically
absent. Importantly, we found that more than 25
percent of the seniors who had been chronically
absent at some point between their 8th grade and
junior year dropped out of high school.
The difference between the first analysis--with
chronic absence in the dropout year included—
and the second analysis with chronic absence in
the dropout year excluded was significantly
different with change in odds ratios of 7.4 and 5.5
respectively. Although the decrease in odds from
7.4 to 5.5 was significant, the drop in odds was
not so great as to suggest that the relationship
between chronic absenteeism and dropout could
be accounted for by absences in that final year.
Using only absenteeism prior to the final year was
still a better predictor for dropping out than any of
the demographic variables of Race, English
Proficiency, Low Income, or Special Education
(with change in odds ratios of 2.7, 4.7, 2.9, and
1.5, respectively).
12th
The Chronic Absenteeism—
Dropout Relationship
From our study and results reported nationally
(Mac Iver and Mac Iver, 2010; Sheldon &
Epstein, 2004), it is apparent that chronic
absenteeism is a strong and early predictor of
dropping out of school. In this section, we
provide an overview of the relationship between
chronic absenteeism and dropping out by using
our longitudinal data set to answer four questions:

How early does chronic absenteeism
predict dropout?

What are the cumulative effects of
chronic absenteeism?

How well does chronic absenteeism
predict dropout independently and in
conjunction with other risk factors?

What is the relationship between chronic
absenteeism, GPA, and dropout?
How Early Does Chronic Absenteeism
Predict Dropout?
To look at chronic absenteeism as an early
indicator of dropout, we used chronic absenteeism
at each year to predict whether a student dropped
Page | 10
out. Using multiple predictors in a single
regression allowed us to look at the independent
effects that being chronically absent in any
particular year had on dropping out. The change
in odds statistics are represented in Figure 8. The
results showed that for each year that a student
was chronically absent, his or her odds of
dropping out approximately doubled. As these are
independent effects, the results for each year can
be thought of as the effect that being absent in that
school year would have if the student was not
chronically absent during any other school year.
For example, the eighth grade result of 2.1 can
be thought of a student being more than twice as
likely to drop out if he or she had been chronically
absent in the eighth grade but not chronically
absent in any subsequent year.
Figure 8. The Change in Odds of Dropping Out
Associated with Chronic Absenteeism in Each
School Year
CA in 12th grade
1.69
CA in 11th grade
2.32
CA in 10th grade
CA in 9th grade
CA in 8th grade
2.70
2.25
2.10
Increase in Odds of Dropping Out
These results show the increased likelihood of
dropping out if a student was chronically absent in
that particular grade, but not chronically absent in
any other grade. The lower statistic in the twelfth
grade is a phenomenon that we saw across
analyses (e.g., see Table 6). That is, chronic
absence in the twelfth grade was not as strongly
related to other predictors and outcomes as it was
in other grades. This anomaly is possibly
accounted for by the overall increase in
absenteeism during the senior year, even in the
most successful students.5
A noteworthy limitation in this analysis is that it is
restrained by the availability of longitudinal data.
At this point, we could only go back as far as
2006 (the eighth grade for students scheduled to
graduate in 2010) to use chronic absenteeism as a
variable for predicting dropout. As more data
become available, we intend to explore the
chronic absenteeism-dropout relationships with
absenteeism data from earlier grade levels.
What are the Cumulative Effects of
Chronic Absenteeism?
What may be implicit from the previous section is
that the risks associated with chronic absenteeism
are cumulative. When the number of years that a
student was chronically absent was used to predict
dropping out, we found that for each year a
student was chronically absent, the odds of
dropping went up 2.21 times, on average. The
actual proportions of Utah students who dropped
out, given the number of years they were
chronically absent, is presented in Table 3.
Table 3. Proportion of Students Dropping Out by
Number of Years the Student was Chronically
Absent
Number of Years
Chronically Absent
0
1
2
3
4
5
Percent Who Dropped
Out
10.3%
36.4%
51.8%
58.7%
61.3%
Not Reported (<1% )
Page | 11
This portrayal of cumulative risk is important for
two reasons. First, it shows a significant increase
in risk of dropping out after only one year of
chronic absenteeism. Second, after two years or
more of being chronically absent, the student is
more likely than not to drop out of school.
How Well Does Chronic Absenteeism
Predict Dropout in Conjunction With
Other Risk Factors?
To understand which variables should be used to
best identify students likely to dropout, we ran
three binary logistic regression models predicting
dropout from different factors.6

Model 1 used the number of years that a
student was chronically absent as the sole
predictor of dropping out.

Model 2 used both the number of years
that a student was chronically absent and
whether the student had a GPA above or
below 1.8 as predictors of dropout.7

Model 3 used years chronically absent,
having a GPA above or below 1.8 and all
of the predictor variables described in the
first section of this brief (i.e., Low
Income, Special Education, English
Proficiency and Racial Minority) as
predictors of dropout.
The results of all three models are presented in
Table 4, which reports the proportion of dropouts
identified by each model.
More than 25 percent of the
seniors who had been
chronically absent at some
point between their 8th grade
and junior year dropped out of
high school.
Table 4. Percent of Dropouts Identified Through
Three Different Models.
Predictors Used
Chronic Absence
GPA
Demographics
Percent of
dropouts identified
Model
1
x
Model
2
x
x
20.6%
59.4%
Model
3
x
x
x
54.6%
Note. Using only the demographic variables (i.e., Low
Income, Special Education, English Proficiency, and
Racial Minority) in a model allowed for the
identification of 6.8 percent of dropouts.
It is clear from the results that the second model,
which uses GPA and Chronic Absence as
predictors was an efficient model. We were not
able to identify any additional students as likely to
dropout by adding the demographic predictors.
Interestingly, some students who eventually did
drop out were identified as likely to drop out
using the second model but identified as not likely
to drop out using the third model.
What is the Relationship between
Chronic Absenteeism, GPA, and
Dropout?
The previous analysis showed both GPA and
chronic absenteeism to be strong predictors of
dropout. To date, the inter-correlations between
all three variables (chronic absenteeism, GPA and
dropout) have been insufficiently considered. To
understand these relationships, we considered
GPA as a mediator of the relationship between
chronic absenteeism and dropout. This is
defensible because the GPA variable was
cumulative GPA, which necessarily came after the
chronic absences and before the dropout. The
gray arrows in Figure 9 illustrate the mediated
relationship in which chronic absenteeism
predicted GPA, which in turn predicted Dropout.
Results from the analysis showed the mediation to
be significant and indicated that GPA partially
mediated the relationship between chronic
absenteeism and GPA.
Page | 12
As shown in Figure 9, the simple correlation
between chronic absence and dropout (represented
by the gold arrow) was r = .44.8 This represented
a medium sized positive relationship between
chronic absenteeism and the likelihood of
dropping out (i.e., students who were chronically
absent were more likely to dropout). After the
mediated relationship was accounted for, the
unmediated correlation between chronic
absenteeism and dropout (represented by the red
arrow) was reduced by 70 percent to a much
smaller but still significant correlation: r = .13.
This result can be interpreted as indicating that 70
percent of the relationship between chronic
absenteeism and dropping out can be accounted
for by the indirect effects (i.e., chronic absence
influences grades, which, in turn, influence
dropping out) and 30 percent of the relationship
between chronic absenteeism and dropping out is
completely independent of GPA.
Figure 9. Direct and Indirect Effects of Chronic Absenteeism on Dropping Out
GPA
0.134497
Chronic
Absence
.443178
Dropout
NOTE: In this figure, the gold arrow represents the simple correlation between chronic absence and dropout. The gray
arrows represent the indirect or mediated effects in which higher chronic absence predicted lower GPA, and then the
lower GPA predicted a higher likelihood of dropping out. The red arrow represents the direct or unmediated effect in
which the relationship between higher chronic absence and higher likelihood of dropping out is independent of GPA.
Page | 13
Conclusion
The findings from this study raise important
considerations for policymakers and practitioners
alike. First, this research emphasizes the need for
early identification of students who are
chronically absent. Qualified school personnel,
such as school counselors, can mitigate the longterm effects of chronic absenteeism through early
identification and intervention. Early
identification provides time to repond to students
who are chronically absent before their absences
impact their persistence or completion of high
school. Knowing when, and for whom, chronic
absence is likely to occur allows for specifically
targeted interventions.
Next, this exploratory study has identified chronic
absenteeism as predicting dropout as early as the
eighth grade. As more years of data become
available, we will be able to look back even
further and, presumably, predict dropout from
earlier grades. Knowing more about the causal
process (e.g., the magnitude of the indirect effect
of chronic absence on dropout as mediated by
GPA) will allow for stronger and more targeted
preventions and interventions. The mediation
model presented here may be thought of as an
early inroad to modeling the causal agents that
result in students dropping out. As more variables
become available the model will become more
sophisticated and, potentially, able to identify
students at-risk for dropping out much earlier.
This also offers a wider range of opportunities for
prevention and intervention programs and
services. In the meantime, however, this research
highlights the need to consider attendance
policies, and how students may be encouraged (or
deterred) from making up their school work when
absent as this impacts their grade point average or
credits earned.
either their junior or senior year. Specifically, 22
percent of the students dropped out in their junior
year and 56 percent of the students dropped out in
their senior year. Thus, again, this research
confirms that the process of dropping out is
protracted. Although the literature shows that
disengagement that results in being pushed out or
dropping out of school begins in the earlier grades
(e.g, Belfanz, Herzog, and Mac Iver, 2007).
Similar to studies nationally, most Utah students
who ultimately dropout persist in their education
through to the later grades (juniors or seniors).
Thus, careful attention is needed to the
experiences of these students, the early indicators,
and the factors leading to their pushout or
departure from school prior to graduation.
For each year that a
student is chronically
absent, his or her odds of
dropping out
approximately double.
Finally, our analysis demonstrated that most
students who dropped out (78 percent) did so in
Page | 14
References
development: The differential effects of school
exposure. Sociology of Education, 83(4) 271-286.
Attwood, G. & Croll, P. (2006). Truancy in
secondary school pupils: Prevalence, trajectories
and pupil perspectives. Research Papers in
Education, 21 (2006): 467–484
Rothman, S. (2001) School absence and student
background factors: A multilevel analysis.
International Education Journal 2(1): 59–68.
Balfanz, R. & Byrnes, V. (2012). Chronic
Absenteeism: Summarizing What We Know From
Nationally Available Data. Baltimore: Johns
Hopkins University Center for Social
Organization of Schools.
Balfanz, R., Herzog, L., & Mac Iver, D.J. (2007).
Preventing student disengagement and keeping
students on the graduation path in urban middlegrade schools: Early identification and effective
interventions. Educational Psychologist, 42, 223235.
Buckner, J. C., Bassuk, E. L., & Weinreb, L. F.
(2001). Predictors of academic achievement
among homeless and low-income housed children.
Journal of School Psychology, 39, 45–69.
Sheldon, S. B., & Epstein, J. L. (2004). Getting
Students to School: Using Family and Community
Involvement to Reduce Chronic Absenteeism.
School and Community Journal 4(2): 39-56.
Utah State Office of Education. 2011-2012 Pointin-time report. Downloaded 6/04/12 from
http://www.schools.utah.gov/fsp/HomelessEducation/Forms-Checklists-Data.aspx
US Department of Education (2002). The
education for homeless children and youth
program: Learning to succeed. Washington, DC:
US Department of Education.
US Department of Education (2004). Education
for homeless children and youth program.
Washington, DC: US Department of Education.
Chang, H. N., & Romero, M. (2008.) Present,
Engaged, and Accounted For: The Critical
Importance of Addressing Chronic Absence in the
Early Grades. Report. New York, N.Y.: National
Center for Children in Poverty.
ECONorthwest. 2011. Chronic Absenteeism in
Oregon: Data Exploration. Portland, OR.
Mac Iver, M. A. & Mac Iver, D. J. (2010).
Gradual disengagement: A portrait of the 2008-09
dropouts in Baltimore city schools. Baltimore
Education Research Consortium.
Moonie S., Sterling D.A., Figgs L.W. & Castro
M. (2008). The relationship between school
absence, academic performance, and asthma
status. Journal of School Health. 78(3): 140–148.
Ready, D. (2010.) Socioeconomic disadvantage,
school attendance, and early cognitive
Page | 15
Footnotes
1
For the analyses describing student level covariates
and the profile of absenteeism we used a crosssectional dataset that included all students who were
enrolled in a Utah public school for at least 20 days
(one month) during the 2010-11 school year. Students
who attended more than one school during the school
year (4.9 percent) were only counted once, using the
data from the school they attended for the most days.
The final data set contained 587,402 K-12 students
from the state of Utah. All data were obtained from the
Utah State Office of Education and were made
available to the UEPC one of the partners in the Utah
Data Alliance. At no point in this analysis were
individual students identified nor will individual
students’ identity be shared.
2
For the analyses that focused on the relationship
between chronic absenteeism and dropout rates, we
used a longitudinal dataset that followed a cohort of
39,141 Utah students who were enrolled in the seventh
grade in 2005 and were scheduled for graduation in
2010. There were concerns about the validity of data
for seventh grade students who exited during their
seventh grade year. We did not see the same problems
in the data from other grades so we eliminated from the
study the 1794 students who did not continue into 8 th
grade. The remaining sample consisted of 37,347
students. Of these students, 75.9 percent (28,339) went
on to completed high school or remained enrolled into
the 2011 school year, 19.2 percent (7169) did not finish
high school, and 4.9 percent (1839) transferred out of
public education with no further information available
concerning them. We excluded from this study the 4.9
percent of students who transferred out of public
education as there is no way of knowing if they
graduated or not. Once the data for students who
transferred were excluded, 35,508 students were
included in the final data set, which accounted for 91
percent of the population.
3
Here we provide a more complete explanation of
how change of odds ratios work. For illustration
purposes of how the change in odds ratio works,
consider family income level as a predictor of chronic
absence. We found that 17 percent of the students
who qualified for free or reduced lunch (a general
measure of family income status) were chronically
absent and that 9.7 percent of the students who did not
qualify were chronically absent. This resulted in .21 to
1 odds of being chronically absent for the students
receiving free and reduced lunch and .11 to 1 odds of
being chronically for students the students who did not.
This difference, reported as a change in odds ratio,
shows that the odds of a student identified as low
income being chronically absent were 1.9 times greater
than the odds of a student not categorized as low
income. This means that a student who received free
and reduced lunch was 90 percent more likely to be
chronically absent than a student who did not receive
free or reduced lunch.
4
A hazard analysis is a common method of assessing
risk over time, particularly when the number of at-risk
units changes over time (as it does in this analysis).
Hazard is the conditional probability that any student
would drop out in a given year, provided he or she had
not dropped out in any previous year. For example,
there were 33,685 students from our sample cohort still
enrolled in 2009 (their junior year) and 1569 of them
dropped out that year. Dividing the number of students
who were enrolled by the number of students who
dropped out (1569/33685 = .0466) showed that about
4.7 percent of the students who were enrolled that year
dropping out that year so the hazard (risk of dropping
out) in 2009 was .047. Hazard was plotted over five
years of data collection for both students who had been
and students who had not been chronically absent,
resulting in hazard functions for each group.
5
This anomaly is informally referred to as “senioritis.”
6
Regression models predicted the log odds {logit(P)
= ln(P/P-1)} of being chronically absence from each
predictor variable in the model yeilding the following
equations:
Model 1: logit(P)=a + b1(Years Chronically
absent)
Model 2: logit(P) = a + b1(Years chronically
absent) + b2 (GPA cut score)
Model 3: logit(P) = a + b1(Years chronically
absent) + b2 (GPA cut score) + b3(Low Income) +
b4(Special Education) + b5(English Proficient) +
b6 (Racial Minority)
Page | 16
Where GPA cut score, Low Income, Special
Education, English Proficiency, and Racial Minority
are all dicotomous variables coded so that 1 equals the
at-risk group (i.e., Cummulative GPA below 1.8,
eligible for free or reduced lunch at any time during
data collection, eligible for special education services
at any time during data collection, eligible for ELL
services at any time during data collection, and nonwhite).
7
The cut score of 1.8 was determined through a
regression process to be the “tipping point” at which
students were best classified as likely to graduate or
likely to drop out.
8
The correlation coefficient, r, measures the direction
and magnitude of the relationship between two
variables. The correlation coefficients reported here,
.44 and .13 are both positive (students who are
chronically absent are more likely to dropout) and
represent medium and small relationships, respectively.
The simple relationship between CA and dropout is
.44—that is the relationship observed by directly
comparing chronic absence status with dropout status.
The mediated relationship is .13—that is the
relationship between CA and dropout after GPA has
been statistically accounted for.
Page | 17
We wish to acknowledge our partnership with the Utah Afterschool Network with whom we
have collaborated through the award of a technical assistance grant from the C.S. Mott
Foundation and our appreciation to Dr. Hedy Chang for her review of this brief. Together we
work increase attention to the value of afterschool and out-of-school time opportunities to
reduce chronic absence.
http://uepc.ed.utah.edu
This report uses data made available through a Data Share Agreement between the USOE and the UEPC. In addition, the
UEPC maintains access to data as a member of the Utah Data Alliance, a partnership of the Utah State Office of Education,
the Utah System of Higher Education, the Utah College of Applied Technology, the Utah Department of Workforce
Services, the Utah Education Policy Center, and the Utah Education Network. The views expressed are those of the authors
and not necessarily the USOE or the Utah Data Alliance partners.
Download